What We Learned Deploying AI within Bloomberg's Engineering Organization

Time: 2:25 PM

Speaker Bio: Head of Technology Infrastructure at Bloomberg Engineering.

Speaker Profile: Full Speaker Profile

Company: Bloomberg is a major financial data and media company deploying AI at enterprise scale.

Focus: Lessons from deploying AI tools across a large engineering org. Practical insights from a mature tech company.

Reference: LinkedIn Profile

Slides

Slide: 14-25

Key Point: The slide predicts five major industry trends toward more specialized, integrated, and usage-based AI implementations with better governance and security built into the models themselves.

Literal Content:

• Title: “Where is the Industry Going?”
• Five trends listed:
  1. LLM-Ready Data Fabric - Agent-readable schemas, lineage, usage hints
  2. Specialist-First Model Grid - Tiny expert models → routed to large generalists on demand
  3. Invisible Copilots - One-click LLM actions inside Slack / CRM / BI
  4. Secure Models from Within - Safety scores from weights & logprobs, not outer wrappers
  5. Usage Based Pricing Models - Fewer workers → Fewer Seats → Licensing focused in usage

Slide: 14-27

Key Point: This slide showcases Bloomberg’s massive scale of operations, emphasizing their workforce, engineering capabilities, and enormous data processing capacity that positions them as a major player in AI/ML applications.

Literal Content:

• Title: “Bloomberg by the numbers”
• Left column (people):
  • Globe icon: 26,000+ employees, worldwide
  • Building icon: 9,000+ engineers
  • News icon: 2,900+ journalists and analysts
  • Database icon: 2,000+ data specialists
  • Robot icon: 400+ employees working on AI and ML applications
• Right column (data):
  • Chart icon: 600+ billion ticks per day, from every asset class & market
  • Building/people icon: 96,000 companies. 2M+ entities. 3.3M bios of executives, leaders & govt officials
  • News icon: 1.5 million news articles ingested per day, from 175,000+ vetted sources
  • Location icon: 100+ Alternative Data sources across multiple sectors
• Footer: “TechAtBloomberg.com” and Bloomberg Engineering logo

Notes

• Lead of AI infrastructure
• 9,000+ engineers, 2000 data specialists, 400 ML/AI people
• Internal stack
  • Largest private network
  • Huge JavaScript framework (tens of millions LoC)
  • Lots of internal libraries being used
• What is AI for coding
  • Usage dropped really quickly once we moved back greenfield
  • “Vibe coding, where 2 engineers can create the tech debt of 50 engineers”
  • What work do our developers not want to do
• Example 1: Uplift agents
  • The patch
  • The rationale
  • Why we’d do it
  • Determine verifiability
• Example 2: Incident response agents
  • Overwhelming number of alerts
  • Hard to find context info
  • Understand contributing factors
• Changing people
  • 20+ training program — well-established training problem
  • Change agent
  • Show how to use it with Bloomberg’s technical
• Individual contributors use it more than leadership
• Changes the cost function of engineering